B-RQL3-markdown-ui-bundle-budget-measured-against-the-playground
State: closed/duplicate
The PLAN budget names “layer + engines add < 150 KB gzip to a product’s main chunk”, but the 314 690 B gzip figure was measured against the playground’s entry, which is not a product. A budget measured against the wrong artifact is worse than a missed budget.
The work is therefore to measure it against a real product’s main chunk — the determined products that consume the layer — and to treat the playground number as an upper bound: it includes surface chrome (find bar, frontmatter panel, hover preview) that not every product loads.
The numbers stand as evidence. After the per-mode split in
#1722, the eager load on edit routes went 674 225 B gzip
to 314 690 B gzip; 131 028 B of that is the Nuxt + @nuxt/ui entry that exists with or without this
layer; the editor and renderer chunks are both behind import() and appear in no route’s eager set.
Splitting the surface chrome is the suggested remedy, explicitly NOT the acceptance condition.